Measure AI tooling.Decide with data.

HubBound Metrics shows what your AI tooling costs, who uses it and how much of what ships carries AI evidence, so spend, adoption and delivery decisions rest on numbers. Boundy AI, its AI agent, answers your questions about them in plain language.

Book a demo

What does HubBound measure?

HubBound measures AI tooling on three levers: adoption (who works with AI, how often and with which tools), spend (what the tokens cost and how much prompt caching saves) and delivery (the commits, pull requests and lines of code that carry evidence of AI assistance). The dashboard groups every metric into eight sections, starting with an Overview of five headline numbers compared with the previous period.

Overview

The headline numbers for the period, and how they moved.

Five headline numbers, each compared with the previous period of the same length:

  • Sessions
  • Tokens
  • Commits
  • Pull requests
  • Lines added

Adoption

Who works with AI, how often, and which tools, models, Skills and MCP servers they reach for.

  • Sessions

    How often the team opens an AI tool, and which ones.

    • Sessions over time
    • Active with AI
    • Activity calendar
    • Sessions by tool
    • Session mix by model

  • Skills & MCP

    The skills and MCP servers the team reaches for.

    • Most used skills
    • Most used MCP servers

Spend

What the usage bills, where the money goes, and how much of it prompt caching keeps off the bill.

  • Tokens

    Token volume, what it was made of, and where it went.

    • Tokens over time
    • What the tokens were
    • Tokens by model
    • Tokens by provider

  • Cost

    What the usage bills, and how much of that is measured.

    • Cost over time
    • Cost by model
    • Cost by person
    • Cost traced to delivery
    • Measurement coverage

  • Cache efficiency

    What prompt caching saves, and how well it works.

    • Cache savings
    • Savings rate over time
    • Cache savings by model
    • How well the cache works

Delivery

How much of the code your team ships carries evidence of AI assistance, and how it moves through review.

  • Commits & pull requests

    Commits and pull requests, and how much AI touched them.

    • Commits over time
    • AI involvement in commits
    • Commits by tool
    • Pull requests by tool
    • Pull request outcomes
    • Time to merge

  • Code lines

    Lines of code landing, and how many carry AI evidence.

    • Lines added over time
    • Lines added by repository
    • Lines added by model

Where does HubBound get its metrics, and who sees them?

HubBound builds its metrics from two sources: the agent telemetry that the HubBound Runtime and CLI report from each developer's machine, and GitHub. An Org Admin sees the whole organization and can narrow the view to one member; every other member sees only their own activity.

  • Agent telemetry

    Reported by the HubBound Runtime and CLI while developers work: sessions, tokens, cost, prompt caching, the tools and models in use, and which Skills and MCP servers were called.

  • GitHub

    Read from your organization's GitHub: commits, pull requests and their outcomes, time to merge, and lines added, each with or without evidence of AI assistance.

Who sees what

  • Org Admin

    Sees the whole organization, and can switch to their own stats or to one member at a time.

  • Team Admin and Member

    See their own activity only. Metrics are not split by team: the view is the organization or one person.

  • Collection switches

    An Org Admin turns Agent metric collection and Github metric collection on or off in Settings, General.

    See the organization settings

  • Plan

    Analytics is part of HubBound's paid plans, and how far back it reaches depends on the plan.

    Compare plans

What decisions can a team make with HubBound Metrics?

HubBound Metrics answers the questions that decide where AI budget, onboarding effort and review time go. Each question below names the metrics that answer it and the decision they support. Where a question reads two metrics side by side, HubBound shows both for the same period; it does not compute a link between them.

  1. Spend

    Where is our AI spend going?

    Read

    • Cost by model
    • Cost over time

    Decide

    Move routine work to a cheaper model where the team agrees the output holds up, then check Cost by model in the next period to confirm the shift.

  2. Spend

    How much of our spend is tied to shipped work?

    Read

    • Cost traced to delivery
    • Measurement coverage

    Decide

    Check coverage first. Spend traced to commits and pull requests is the part you can defend in a budget review; look into the rest before you cut it.

  3. Spend

    Is prompt caching lowering the bill?

    Read

    • Cache savings
    • How well the cache works
    • Cache savings by model

    Decide

    Keep the setups where caching absorbs a real share of the bill, and review how sessions run on the models where it saves little.

  4. Adoption

    Who is using AI, and who has not started?

    Read

    • Active with AI
    • Activity calendar
    • Insights: Dormant members

    Decide

    Aim onboarding where use is thin: distribute a starter Kit to the teams that need it, then watch Active with AI over the next period.

  5. Adoption

    Which AI tools does the team actually use?

    Read

    • Sessions by tool
    • Commits by tool
    • Pull requests by tool

    Decide

    Keep paying for the tools the team works in and question the seats nobody opens. A HubBound Kit installs into whichever supported tools you keep.

  6. Delivery

    How much of what we ship carries AI evidence, and is it merging?

    Read

    • AI involvement in commits
    • Lines added over time
    • Pull request outcomes
    • Time to merge

    Decide

    Set a baseline for AI's share of delivery. If Time to merge climbs while AI involvement rises, read the two side by side and look at review capacity before pushing adoption further.

  7. Governance

    Are the Skills and MCP servers we rolled out being used?

    Read

    • Most used skills
    • Most used MCP servers

    Decide

    Keep what the team reaches for. Rewrite or retire the Skills and MCP servers that never show up, and ship the change as a new Kit version.

What does the HubBound metrics dashboard show?

The HubBound metrics dashboard has two tabs, Insights and Data. Data holds the metric grid: an Overview of five headline numbers with their change against the previous period, then seven sections from Sessions to Skills & MCP. Pick a section below to see what it holds.

Your organization
Illustrative sample

Sessions

+9.6% Compared with the previous period

Tokens

+14.1% Compared with the previous period

Commits

+7.3% Compared with the previous period

Pull requests

+11.2% Compared with the previous period

Lines added

+6.8% Compared with the previous period

Sessions over time

AI sessions opened each day.

+9.6%

Active with AI

Distinct people who used AI at least once this period.

Sessions by tool

Which AI tools your team actually opens.

  • Claude Code612
  • Cursor388
  • Codex196
  • Copilot88

Illustrative sample data. The layout mirrors the Data tab of HubBound's analytics dashboard; model names are examples.

  • Filter the period and the scope

    Pick 7 days, 30 days, 3 months or 12 months, then narrow by member, tool or model. Every number compares with the previous period of the same length.

  • See the right slice

    Org Admins read the organization, their own stats or one member. Everyone else reads their own activity.

  • Arrange it your way

    Customize reorders or hides sections, and each section collapses. The arrangement is saved in the viewer's browser.

  • Export a PDF report

    Export PDF turns the current view into a report that follows your arrangement, ready for a budget review or a leadership update.

What are HubBound Insights?

Insights is the HubBound dashboard tab that reads the period for you: up to eight findings about adoption, delivery, cost and activity, each with a written headline, a severity, a confidence level and the evidence behind it, compared with the previous period. A finding without enough activity behind it is held back and named instead of guessed.

  • Eight findings

    Adoption rate, Dormant members, Adoption concentration, AI delivery footprint, Repository spread, Cost concentration, Model mix and Activity change.

  • Ranked by severity

    Needs attention first, then Working well, then For context. A finding with a thin sample carries a Low confidence badge.

  • Organization or personal

    An Org Admin reads the organization, and findings that name people appear only to them. Everyone else reads their own activity. The tab takes a period; tool and model filters belong to Data.

Illustrative sample. Finding titles and labels come from the app; headlines are examples.

What is Boundy AI?

Boundy AI is the AI agent inside HubBound Metrics: ask about your organization's AI usage, adoption or costs in plain language, and it answers with KPI cards, charts, tables and leaderboards built from your own data.

Illustrative sample. Boundy AI's labels come from the app; the conversation and its numbers are examples.

  • Ask, don't filter

    Type the question the way you would ask a colleague. Boundy AI picks the chart that answers it.

  • Same access as the dashboard

    Rankings that name people are for Org Admins only; everyone else reads their own activity.

  • One step from Ink

    With Memories on, Boundy AI links to Ink for what your team decided, not what it used.

    See Memories, Projects and Ink

  • Billed past a threshold

    Core includes Boundy AI questions up to a usage threshold; past it, they are billed on demand.

    See pricing

How do HubBound Metrics close the loop on a Kit rollout?

HubBound Metrics does not attribute results to a Kit, but it shows what moved after one ships. Because every number is compared with the previous period of the same length, reading one full period after a rollout gives you the before and the after.

  1. Ship the standard

    Distribute a Kit to a team or to the whole organization from Org & Teams.

  2. Let one period pass

    Wait a full period, for example 30 days, so the comparison with the previous 30 days spans the rollout.

  3. Read what the Kit touches

    Find the Kit's Skills and MCP servers in Most used skills and Most used MCP servers, with AI involvement in commits and Active with AI beside them. Usage is counted per Skill and MCP server, not per Kit, and the view is the organization or one member, not a team.

  4. Adjust and version

    Keep, rewrite or retire what the numbers point to. The change saves a new Kit version, and teams that follow the latest version receive it.

See how a Kit is distributedSee how Kits are versioned

See how it all connects.

Each piece of HubBound builds on the others. Pick where you want to go next.

Artifacts & Kits

The standards you build, package and distribute.

Explore

Home

The big picture: what HubBound is and how it works.

Explore

CLI

One binary that puts it all on every machine.

Explore

Security

On-demand Skill audits, with a verdict and a fix for every finding.

Explore

Org & Teams

How your organization and teams are structured.

Explore

Live Planning

Your agent's plan, shared, edited live and reviewed before it runs.

Explore

Memories, Projects and Ink

One bundle: your team's shared context, and Ink to ask it anything.

Explore

Put numbers behind every AI decision.

See how HubBound measures adoption, spend and delivery across your organization, and what your team can decide with it.

Book a demo